On Mon, May 13, 2013 at 04:44:13PM -0500, steven.kin...@amd.com wrote:
> From: Steven L Kinney <steven.kin...@amd.com>
>
> Implement a perf PMU to handle IOMMU PC perf events. This PMU will handle
> static counter perf events relative to the AMD IOMMU Performance Counters.
>
> To invoke the AMD IOMMU PMU issue a perf tool command such as:
>
> ./perf stat -e amd_iommu/config=<config-data>,config1=<config1-data>/u
> <command>
>
> For example:
>
> ./perf stat -e amd_iommu/config=0x8000000000000005,config1=0/u <command>
>
> The resulting count will be how many IOMMU total peripheral memory
> operations were performed during the command execution window.
>
> Signed-off-by: Steven Kinney <steven.kin...@amd.com>
> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpa...@amd.com>
> ---
> +PMU_FORMAT_ATTR(iommu_raw, "config:63");
> +static int perf_iommu_init(struct perf_event *event)
> +{
> + struct hw_perf_event *hwc = &event->hw;
> + struct perf_iommu *perf_iommu;
> + u64 config, config1;
> +
> + /* initialize the cntr_assign_mask */
> + cntr_assign_mask = 0;
> +
> + /* test the event attr type check for PMU enumeration */
> + if (event->attr.type != event->pmu->type)
> + return -ENOENT;
> +
> + perf_iommu = &perf_iommu_fetch;
> +
> + if (perf_iommu) {
> + config = event->attr.config;
> + config1 = event->attr.config1;
> + } else
> + return -EINVAL;
> +
> + if (event->pmu != &perf_iommu->pmu)
> + return -ENOENT;
> +
> + /* make sure the event->attr.config is raw (vendor specific) */
> + if (!(config & IOMMU_PC_RAW_CONFIG_MASK))
> + return -EINVAL;
Is this really needed? Non of the uncore PMU drivers have a magic config bit
either. Seems weird to me.
> + /* integrate with iommu base devid (0000), assume one iommu */
> + perf_iommu->max_banks =
> + amd_iommu_pc_get_max_banks(IOMMU_BASE_DEVID);
> + perf_iommu->max_counters =
> + amd_iommu_pc_get_max_counters(IOMMU_BASE_DEVID);
> +
> + /* update the hw_perf_event struct with the iommu config data */
> + hwc->config = config;
> + hwc->extra_reg.config = config1;
> +
> + return 0;
> +}
Also it might make sense to add a few event aliases to both this IOMMU and the
AMD uncore driver.
Search for uncore_.*_events in arch/x86/kernel/cpu/perf_event_intel_uncore.c
for examples.
This way you can also make the IOMMU a tad more useful by providing some events
you think are the most useful for people interested in this thing.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
